Semiparametric Construction of Spatial Generalized Hedonic Models for Private Properties

This paper analyzes the empirical hedonic prices for non-standard condominiums and single family houses using nonparametric estimates as well as a generalized additive model for the spatial generalization of the attractiveness of all Swiss communities. We find that the assumption of log-linearity for continuous variables does not hold but can be replaced by partwise log-linear or quadratic terms. Due to the topographical segmentation of Switzerland, driving times seem to be more adequate than geographical distances to explain the price level of a village with the price level of its neighbours. We show, that using metric multidimensional scaling, the driving times between the villages can be converted into artificial coordinates with three principal axis to serve as a basis for the prediction of the macro-locations of the Swiss villages.